Machine learning model could accelerate drug discovery

Researchers from the University of Warwick in the United Kingdom have developed a machine learning model capable of predicting the interactions between proteins and drug molecules with 99 percent accuracy. Findings have been published in Science Advances.

The algorithm aims to accelerate drug discovery by using only a handful of reference experiments to provide accurate results determining whether a drug molecule will bind to a target protein. Additionally, the model is able to provide insight into intermolecular forces by outlining solutions to material-science problems like modeling to silicon surfaces.

"This work is exciting because it provides a general-purpose machine learning approach that is applicable both to materials and molecules,” said James Kermode, from the University of Warwick's Center for Predictive Modeling. "The research is expected to lead to a significant increase in the accuracy and transferability of models used for drug design and to describe the mechanical properties of materials."

Researchers developed the algorithm by combining local information of each atom in a structure and makes it applicable on different classes of chemical, material science and biochemical problems. This design makes the machine learning model able to predict the stability of organic molecules and the energy balance in silicone structures for microelectronic applications.